CN113907880B

CN113907880B - Positioning device and method of interventional device, computer device and readable storage medium

Info

Publication number: CN113907880B
Application number: CN202010647212.4A
Authority: CN
Inventors: 王彦磊; 王心怡; 孙毅勇
Original assignee: Shanghai Microport EP MedTech Co Ltd
Current assignee: Shanghai Microport EP MedTech Co Ltd
Priority date: 2020-07-07
Filing date: 2020-07-07
Publication date: 2023-05-16
Anticipated expiration: 2040-07-07
Also published as: CN113907880A; WO2022007815A1

Abstract

The invention provides a positioning device and a positioning method of an interventional device, and a computer device and a storage medium. In a training stage, at each sampling time, calculating the spatial position information of the second locus according to the spatial position and direction information of the first locus and the spatial distance between the first locus and the second locus, and forming a first V-P data pair by the voltage information of the second locus relative to a reference position and the spatial position information of the second locus, so as to train a neural network model and obtain an initial VP model for describing the mapping relation between the voltage and the locus; and in the positioning stage, at each sampling time, calculating the spatial position of the third site by using the initial VP model according to the voltage information of the third site relative to the reference position. The scheme of the invention can locate and track the intervention device in the target object, and improve the locating accuracy.

Description

Positioning device and method of interventional device, computer device and readable storage medium

Technical Field

The invention belongs to the technical field of medical equipment, and particularly relates to positioning equipment and method of an interventional device, an interventional operation system, computer equipment and a readable storage medium.

Background

There are many medical applications in which it is desirable to determine the exact location of various medical devices, such as catheters and implants, within the body. There are a number of positioning systems developed in current medical devices. Generally, various energy fields such as a magnetic field, an electric field or an ultrasonic field are applied outside the body, and simultaneously, a sensor capable of sensing the corresponding energy field such as a magnetic field sensor, an electric field sensor, an ultrasonic transducer and the like is assembled on a device to be positioned in the body, and then the energy field information sensed by the sensor is converted into position information. Wherein the positioning system based on the magnetic field is mature and has higher precision; but has the disadvantage that the magnetic field sensor is relatively costly and unsuitable for mass use. Therefore, the magnetic field positioning is generally combined with other positioning technologies, and the cost is controlled while the sufficient positioning precision can be ensured.

In the prior art, the following three positioning methods exist:

1. placing a plurality of patch electrodes on the body surface, using electrodes on a catheter in the body to emit current excitation, measuring impedance values between the catheter electrodes and the patches on each body surface, and simultaneously providing accurate position information by a magnetic field sensor assembled on the catheter; after the conversion relation between the impedance values of the plurality of positions and the corresponding positions is calculated, the positioning and tracking of the impedance values of the common electrode can be realized only by the impedance values of the common electrode.

2. The first and second coordinate systems are defined based on the electric field and the magnetic field respectively, the magnetic field sensor on the catheter is close to the electrode, and after the position of the magnetic field sensor in the first coordinate system is related with the position of the magnetic field sensor in the second coordinate system by using a mapping function, the position of the electrode is correspondingly adjusted and calibrated according to the mapping position of the magnetic field sensor in the first coordinate system.

3. Three mutually orthogonal signals are applied to the region of interest of the human body, the voltages between the electrodes on the catheter and the reference electrode in the human body are measured, and the position coordinates of the catheter electrodes can be calculated by using the three voltages in 3 orthogonal directions (namely, xyz three directions).

However, in the above methods, the positioning is based on the electric field, the accuracy is not high, and nonlinear distortion occurs in the electric field in vivo, and this distortion can cause difficulty in accurately modeling the electric field in vivo by using the existing linear model, and further cause inaccurate positioning results.

Disclosure of Invention

The invention aims to provide a positioning device, a positioning method, an interventional operation system, a computer device and a readable storage medium of an interventional device, so as to position and track the interventional device in a target object and improve the positioning accuracy. The specific technical scheme is as follows:

In order to achieve the above object, the present invention provides a positioning apparatus of an interventional device comprising a learning means and an application means, the learning means and the application means being adapted to be placed in a target area of a target object;

the positioning device includes: the device comprises a magnetic field generation unit, an excitation control unit, a data acquisition unit and a processor unit;

the magnetic field generating unit is used for generating a magnetic field passing through the target area;

the excitation control unit is used for applying excitation to at least three electrode patches arranged on the surface of the target object so as to apply an N-axis electric field in the target object, wherein N is more than or equal to 3;

the data acquisition unit is used for synchronously acquiring magnetic field intensity information of a first position point on the learning appliance and voltage information of a second position point on the learning appliance relative to a reference position in all excitation states in a training stage; and, in a positioning phase, synchronously acquiring voltage information of a third site on the application tool relative to the reference position in all excitation states;

the processor unit is configured to calculate, at each sampling time, spatial position information of the second site according to spatial position and direction information of the first site and spatial distance between the first site and the second site, and form a first type V-P data pair from voltage information of the second site relative to the reference position and spatial position information of the second site in different excitation states, where the first type V-P data pair is used to train a neural network model to obtain an initial VP model for describing a mapping relationship between voltage and position, where the spatial position and direction information of the first site are calculated according to magnetic field strength information of the first site; and in the positioning stage, at each sampling time, calculating the spatial position of the third site by using the initial VP model according to the voltage information of the third site relative to the reference position under different excitation states.

Optionally, in the positioning device, the neural network model includes an input layer, a plurality of hidden layers and an output layer that are sequentially connected, the number of neurons of the input layer is equal to the number of different excitation states, and the number of neurons of the output layer is equal to 3.

Optionally, in the positioning device, the processor unit is specifically configured to train to obtain the initial VP model in the following manner:

taking the voltage information in the first V-P data pair as the input of the neural network model, and calculating the spatial position information predicted by the output layer;

calculating an objective function value of the neural network model according to the spatial position information in the first V-P data pair and the spatial position information predicted by the output layer;

and optimizing model parameters of the neural network model according to the objective function value until a preset training ending condition is met, so as to obtain the initial VP model.

Optionally, in the positioning device, the application apparatus further includes a fourth site and a fifth site thereon;

the data acquisition unit is further used for synchronously acquiring magnetic field intensity information of the fourth site and voltage information of the fifth site relative to the reference position in all excitation states in a positioning stage;

The processor unit is further configured to calculate, at each sampling time, spatial position information of the fifth site according to spatial position and direction information of the fourth site and spatial distance between the fourth site and the fifth site, and form a second class V-P data pair from voltage information of the fifth site relative to the reference position and the spatial position information of the fifth site in different excitation states, so as to train to obtain a new VP model; the spatial position and direction information of the fourth site is calculated according to the magnetic field intensity information of the fourth site.

Optionally, in the positioning apparatus, the application device and the learning device are the same device, the fourth site and the first site are the same site, and the fifth site and the second site are the same site.

Optionally, in the positioning device, the processor unit is specifically configured to update the initial VP model according to the second type V-P data pair, so as to obtain a new VP model.

Optionally, in the positioning device, the processor unit is specifically configured to train a neural network model according to the second type V-P data to obtain a new VP model.

Optionally, in the positioning device, the processor unit is specifically configured to periodically train a neural network model according to the second type V-P data within a last preset duration, so as to obtain a new VP model.

Optionally, in the positioning device, the processor unit is further configured to combine the voltage information of the third site relative to the reference position and the spatial position information of the third site in different excitation states into a third class V-P data pair, screen an effective data pair from the third class V-P data pair, and update the initial VP model according to the effective data pair, so as to obtain a new VP model.

Optionally, in the positioning device, the neural network model is selected from one of an error back propagation neural network, a radial basis function neural network, and a support vector machine neural network.

Optionally, in the positioning device, the excitation control unit applies excitation to the at least three electrode patches and performs continuous cyclic high-speed switching between all excitation states, and the data acquisition unit acquires voltage information of the second site and the third site relative to the reference position in all excitation states; or alternatively, the first and second heat exchangers may be,

The excitation control unit applies excitation to the at least three electrode patches at the same time, but the frequencies of the applied excitation are different from each other, and the data acquisition unit acquires the voltage information of the second site and the third site relative to the reference position and performs filtering processing to acquire the voltage information of the second site and the third site relative to the reference position under all excitation states;

the excitation control unit applies excitation to the at least three electrode patches at the same time, but the applied excitation frequencies are different, the data acquisition unit acquires voltage information of the second site or the third site relative to the reference position, and the processor unit performs filtering processing on the voltage information of the second site and the third site relative to the reference position acquired by the data acquisition unit to acquire the voltage information of the second site and the third site relative to the reference position under all excitation states.

Optionally, in the positioning device, the excitation is constant current excitation or constant voltage excitation.

Optionally, the positioning device further comprises a communication control unit, which is used for connecting the processor unit with the magnetic field generating unit, the excitation control unit and the data acquisition unit so as to control communication and data transmission between the processor unit and the magnetic field generating unit, the excitation control unit and the data acquisition unit.

Optionally, the positioning device further comprises a display unit, which is in communication connection with the processor unit and is used for displaying the position, the direction, the shape and/or the movement track of the application tool in the target object, wherein the position, the shape, the direction and/or the movement track of the application tool in the target object are determined by the processor unit according to the positioning result of the third site.

Optionally, in the positioning device, a magnetic field sensor is disposed at the first location, and the data acquisition unit acquires magnetic field intensity information of the first location through the magnetic field sensor;

the data acquisition unit acquires voltage information of the second site and the third site relative to the reference position through the voltage sensors;

the interventional device, the learning device and the application device are all interventional catheters.

Optionally, in the positioning device, the different excitation states include M effective excitation states selected from all excitation states, where M is 3-N; or (b)

The different excitation states include all of the excitation states.

Based on the same inventive concept, the present invention also provides an interventional procedure system comprising a positioning device of an interventional device as described above and the interventional device.

Based on the same inventive concept, the present invention also provides a method of positioning an interventional device comprising a learning means and an application means, the learning means and the application means being adapted to be placed in a target area of a target object;

the method comprises the following steps:

in a training stage, receiving magnetic field intensity information of a first site on the learning appliance, which is synchronously acquired at each sampling time, voltage information of a second site on the learning appliance relative to a reference position in all excitation states, calculating the spatial position information of the second site according to the spatial position and direction information of the first site and the spatial distance between the first site and the second site, and forming a first V-P data pair by the voltage information of the second site relative to the reference position and the spatial position information of the second site in different excitation states, so as to train a neural network model, and obtaining an initial VP model for describing a mapping relation between voltage and position, wherein the spatial position and the direction information of the first site are calculated according to the magnetic field intensity information of the first site;

And in a positioning stage, receiving voltage information of a third site on the application tool relative to the reference position in all excitation states, which is synchronously acquired at each sampling time, and calculating the spatial position of the third site by using the initial VP model according to the voltage information of the third site relative to the reference position in different excitation states.

Optionally, in the positioning method, the neural network model includes an input layer, a plurality of hidden layers and an output layer that are sequentially connected, the number of neurons of the input layer is equal to the number of different excitation states, and the number of neurons of the output layer is equal to 3.

Optionally, in the positioning method, the initial VP model is trained in the following manner:

Optionally, in the positioning method, the application device further comprises a fourth site and a fifth site;

the method further comprises the steps of:

in a positioning stage, receiving magnetic field intensity information of the fourth site synchronously acquired at each sampling time, voltage information of the fifth site relative to the reference position in all excitation states, calculating the spatial position information of the fifth site according to the spatial position and direction information of the fourth site and the spatial distance between the fourth site and the fifth site, and forming a second V-P data pair by the voltage information of the fifth site relative to the reference position and the spatial position information of the fifth site in different excitation states for training to obtain a new VP model; the space position and direction information of the fourth site is calculated according to the magnetic field intensity information of the fourth site.

Optionally, in the positioning method, the application device and the learning device are the same device, the fourth site and the first site are the same site, and the fifth site and the second site are the same site.

Optionally, in the positioning method, a new VP model is trained as follows:

And updating the initial VP model according to the second class V-P data pair to obtain a new VP model.

Optionally, in the positioning method, a new VP model is trained as follows:

and training a neural network model according to the second class V-P data to obtain a new VP model.

Optionally, in the positioning method, training a neural network model according to the second class V-P data to obtain a new VP model includes:

and training a neural network model according to the second type V-P data in the latest preset time period periodically to obtain a new VP model.

Optionally, in the positioning method, the method further includes:

and forming a third V-P data pair by the voltage information of the third site relative to the reference position and the space position of the third site under different excitation states, screening effective data pairs from the third V-P data pair, and updating the initial VP model according to the effective data pairs to obtain a new VP model.

Optionally, in the positioning method, the neural network model is selected from one of an error back propagation neural network, a radial basis function neural network and a support vector machine neural network.

Optionally, in the positioning method, the method further includes:

and determining the position, the shape, the direction and/or the movement track of the application tool in the target object according to the positioning result of the third site, and driving a display unit to display the position, the direction, the shape and/or the movement track of the application tool in the target object.

Optionally, in the positioning method, a magnetic field sensor is arranged at the first site, and the data acquisition unit acquires magnetic field intensity information of the first site through the magnetic field sensor;

the second site and the third site are provided with voltage sensors, and a data acquisition unit acquires voltage information of the second site and the third site relative to the reference position through the voltage sensors;

Optionally, in the positioning method, the different excitation states include M effective excitation states selected from all excitation states, where M is 3-N; or (b)

The different excitation states include all of the excitation states.

Based on the same inventive concept, the present invention also provides a computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, implements the steps of the method of localization of an interventional device as described above.

Based on the same inventive concept, the invention also provides a computer device comprising a memory and a processor, the memory storing a computer program, the processor executing the steps of the method of positioning an interventional device as described above when the computer program is executed.

Compared with the prior art, the positioning equipment, the method, the interventional operation system, the computer equipment and the readable storage medium of the interventional device have the following beneficial effects:

according to the invention, the V-P data pair for training is obtained from the first site and the second site on the learning appliance of the interventional device, the initial VP model for describing the mapping relation between the voltage value and the position coordinate value is trained, and then the third site on the application appliance is positioned through the initial VP model, so that the positioning and tracking of the application appliance are realized. Compared with the prior art, the nonlinear conversion model between the electric field and the space position is constructed by adopting the nonlinear fitting characteristic of the neural network technology, so that the problem of low positioning accuracy caused by the nonlinear distortion effect of the electric field during electric field positioning can be solved, and the positioning accuracy is improved.

Drawings

In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic view of an operation module of a positioning apparatus of an interventional device according to an embodiment of the present invention;

FIG. 2 is a flow chart of the use of the positioning device shown in FIG. 1;

FIG. 3 is a flow chart of a positioning algorithm implementation;

FIG. 4a is a schematic view of an annular duct;

FIG. 4b is a view of the placement of various sensors (sites) in the catheter shown in FIG. 4 a;

FIG. 4c is a schematic view of the positioning result of the catheter of FIG. 4a at a certain time;

fig. 5 is a structural diagram of a BP neural network.

Detailed Description

The following describes the positioning device, method, interventional operation system, computer device and readable storage medium of the interventional device according to the present invention in further detail with reference to the drawings and the detailed description. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the drawings are in a very simplified form and are all to a non-precise scale, merely for the purpose of facilitating and clearly aiding in the description of embodiments of the invention. For a better understanding of the invention with objects, features and advantages, refer to the drawings. It should be understood that the structures, proportions, sizes, etc. shown in the drawings are for illustration purposes only and should not be construed as limiting the invention to the extent that any modifications, changes in the proportions, or adjustments of the sizes of structures, proportions, or otherwise, used in the practice of the invention, are included in the spirit and scope of the invention which is otherwise, without departing from the spirit or essential characteristics thereof.

The core idea of the invention is to provide a positioning device, a method, an interventional operation system, a computer device and a readable storage medium of an interventional device, construct a nonlinear conversion model between an electric field and a spatial position based on nonlinear fitting characteristics of a neural network method so as to position and track the interventional device in a target object, and solve the problem of positioning the interventional device in a body. The interventional device comprises a learning means and an application means, the learning means and the application means being adapted to be placed within a target area of a target object.

The positioning device provided by the invention comprises: the device comprises a magnetic field generating unit, an excitation control unit, a data acquisition unit and a processor unit.

Specifically, the magnetic field generating unit is configured to generate a magnetic field that passes through the target area, and at least three electrode patches are disposed on the surface of the target object;

the excitation control unit is used for applying excitation to the at least three electrode patches so as to apply an N-axis electric field in the target object, wherein N is more than or equal to 3;

the data acquisition unit is used for synchronously acquiring magnetic field intensity information of a first position on the learning appliance and voltage information of a second position on the learning appliance relative to a reference position under all excitation states in a training stage; and, in a positioning phase, synchronously acquiring voltage information of a third site on the application tool relative to the reference position in all excitation states;

The processor unit is configured to calculate, at each sampling time, spatial position information of the second site according to spatial position and direction information of the first site and spatial distance between the first site and the second site, and form a first type V-P data pair from voltage information of the second site relative to the reference position and spatial position information of the second site in different excitation states, so as to train a neural network model, and obtain an initial VP model for describing a mapping relationship between the voltage and the position, where the spatial position and direction information of the first site are calculated according to magnetic field strength information of the first site; and in the positioning stage, at each sampling time, calculating the spatial position of the third site by using the initial VP model according to the voltage information of the third site relative to the reference position under different excitation states.

The positioning equipment is divided into a training stage and a positioning stage when in use, training data of a neural network model is acquired by adopting a learning appliance in the training stage, an initial VP model for describing a mapping relation between voltage and position is obtained by training, at least two sites including a first site and a second site are arranged on the learning appliance, the spatial distance between the first site and the second site is known, the first site is positioned by a magnetic field, and the position of the second site can be obtained according to voltage information and the position of the first site. And then, the position of a third site on the application tool is obtained based on the initial VP model in the positioning stage, so that the positioning of the application tool is realized, and the position, the direction and the shape information of the application tool in the target area of the target object can be further determined. Specifically, the first site is provided with a magnetic field sensor, the second site and the third site are provided with voltage sensors, so that the data acquisition unit can acquire magnetic field intensity information of the first site through the magnetic field sensor, and acquire voltage information of the second site and the third site relative to a reference position through the voltage sensors.

In the invention, the neural network model comprises an input layer, a plurality of hidden layers and an output layer which are sequentially connected, wherein the number of neurons of the input layer is equal to the number of different excitation states, and the number of neurons of the output layer is equal to 3.

The processor unit may train to obtain the initial VP model in the following manner:

According to the positioning equipment of the interventional device, the V-P data pair for training is obtained through the first site and the second site on the learning appliance, the initial VP model for describing the mapping relation between the voltage value and the position coordinate value is trained, and then the third site on the application appliance is positioned through the initial VP model, so that the positioning of the application appliance is realized. Compared with the prior art, the method and the device have the advantages that the nonlinear conversion model between the electric field and the space position is constructed by using the nonlinear fitting characteristic of the neural network technology, the problem of low positioning accuracy caused by the nonlinear distortion effect of the electric field during electric field positioning can be solved, and the positioning accuracy is improved.

The following describes in detail the positioning device of the interventional device provided by the invention in connection with fig. 1-5.

Referring to fig. 1, fig. 1 schematically illustrates a working module of a positioning apparatus of an interventional device according to an embodiment of the present invention.

The target object 100 is an application object of the positioning device, and may be a person, an animal or other suitable conductor object. The target region 101 represents the region within the target object 100 where the distal end of the interventional device 105 is located, e.g. the heart chamber.

The interventional device 105 may be a variety of medical catheters, such as electrophysiological catheters, multi-limb catheters for ablation or mapping, or looped catheters, as well as other intracorporeal implant devices. When the interventional device 105 is an electrophysiology catheter, it is typically used for diagnosis or treatment, such as electrocardiographic mapping or delivery of ablation energy, etc. The interventional device 105 shown in fig. 1 is an electrophysiological catheter that typically reaches the target region 101, such as the heart chamber, through a vascular access. The distal end of the interventional device 105 may be fixed or bendable; the portion of the interventional device 105 left outside the body typically has an operating handle 106, and the physician can control the shape or orientation of the distal end of the interventional device 105 by manipulating the handle 106.

The interventional device 105 is equipped with two types of sensors, including a magnetic field sensor 103 and a voltage sensor 104 (i.e. a common electrode, hereinafter referred to as "electrode 104"), i.e. the magnetic field sensor 103 is equipped at a first site on the interventional device and the electrodes 104 are equipped at a second site and a third site. Signals sensed by the magnetic field sensor 103 and the voltage sensor 104 are transmitted to the data acquisition unit 30 through wires inside the interventional device 105. There may be one to more of the magnetic field sensor 103 and the electrode 104.

Depending on the function of the interventional device 105 during positioning, the interventional device 105 may comprise a learning tool, on which the magnetic field sensor 103 and the electrode 104 have to be fitted simultaneously, and an application tool, on which only the electrode 104 has to be fitted (to save costs). Of course, this is merely a functional distinction, and in practical applications the application device and the learning device may be the same device, and the interventional device 105 is exemplified by a medical catheter, and the interventional device may comprise two catheters, namely a learning catheter (i.e. learning device) and an application catheter (i.e. application device), respectively, and the interventional device may comprise only one catheter, which may be used as both a learning catheter and an application catheter. Of course, in other embodiments, the interventional device may also include at least three catheters, one or more of which may be used as a learning catheter, and the others may be used as application catheters, as appropriate, and the invention is not limited in this regard.

The following description will take the interventional device 105 as an example of a medical catheter. In the following description, for convenience, it is assumed that a medical catheter is used as both a learning catheter and an application catheter, with a magnetic field sensor 103, a known electrode 104 (hereinafter 104S) at a fixed distance from the magnetic field sensor 103, and a plurality of other electrodes 104 (hereinafter 104T) to be positioned. The distance is the Euclidean distance (Euclidean Distance) in three-dimensional space. Thus, the position information of the electrode 104S can be calculated according to the spatial position and direction information of the magnetic field sensor 103, and the position information of the electrode 104T needs to be obtained through the neural network model described in the present invention. It should be noted that, the first position on the medical catheter according to the present invention is the position of the magnetic field sensor 103, the second position is the position of the electrode 104S, and the third position is the position of the electrode 104T, so that calculating the spatial positions of the second position and the third position is calculating the spatial positions of the electrode 104S and the electrode 104T.

The magnetic field sensor 103 generally comprises several coils for sensing the strength of the magnetic field at its location, which magnetic field is excited by the magnetic field generating unit 10. The processor unit 40 may calculate the spatial position information and the direction information of the magnetic field sensor 103 based on the magnetic field intensity information sensed by the magnetic field sensor 103, and the specific calculation method belongs to the category of magnetic positioning technology, and reference may be made to the prior art, which is not described herein. The spatial position information generally refers to three coordinate values of x, y and z in a three-dimensional cartesian coordinate system. The direction information is a direction vector of the magnetic field sensor 103, and more generally, a unit direction vector thereof.

The electrode 104 (i.e., the voltage sensor) is a biocompatible metal, such as platinum iridium alloy, gold, and the like. The electrode 104 is used to measure the voltage at its location relative to some reference location. The reference location (not shown in fig. 1) is typically a relatively stable location, and may be selected from a body surface location, or may be selected from a body location (e.g., the coronary sinus). Preferably, a reference electrode is placed at a reference location (e.g., the coronary sinus). In addition, in addition to being used to measure voltage, the electrode 104 may also be used to measure electrophysiological signals or to release ablation energy.

The magnetic field generating unit 10 is for generating an electromagnetic field, and is typically fixed near the target object 100. So that the electromagnetic field 107 it generates can pass through the target area 101. The electromagnetic field may be sensed by the magnetic field sensor 103 to locate the spatial position and orientation of the magnetic field sensor 103 within the body.

The number of electrodes to be attached to the surface of the target is generally not less than 3 (hereinafter, simply referred to as "electrode patches 102" for distinguishing the electrodes 104 on the medical catheter from the above-mentioned electrodes). The excitation control unit 20 may apply a specific electric field to the human body through any two electrode patches 102, that is, apply excitation, preferably apply a constant current or constant voltage signal; preferably, an applied current flows through the target area 101. The excitation control unit 20 selects a plurality of pairs of electrode patches 102 to apply excitation, and the data acquisition unit 30 acquires the voltage value of each electrode 104 on the medical catheter relative to the reference position for subsequent positioning calculation under all excitation states at the same sampling time. The same sampling time refers to a time when the magnetic field intensity data and the voltage information of the magnetic field sensor 103 are collected, that is, the data collecting unit 30 needs to collect the magnetic field intensity data and the voltage information at the same sampling time. The excitation state refers to a state in which the excitation control unit 20 applies a current to the human body through one of the two electrode patches 102 (a pair of electrode patches), and may be also referred to as an excitation axis, and when the number of excitation axes is N, the excitation control unit 20 may be considered to apply an N-axis electric field to the N pairs of electrode patches, thereby realizing the application of an N-axis electric field to the target object. The position of the electrode patch 102 against the body surface is generally fixed throughout the procedure.

In order to obtain the voltage information of the second site and the third site (i.e. each electrode 104) relative to the reference position in all excitation states at the same sampling instant, the following method may be adopted: the excitation control unit 20 applies excitation to the at least three electrode patches 102 and performs continuous cyclic high-speed switching between all excitation states, and the data acquisition unit 30 acquires voltage information of the second site or the third site (i.e., each electrode 104) with respect to a reference position in all excitation states; or alternatively, the first and second heat exchangers may be,

the excitation control unit 20 applies excitation to the at least three electrode patches 102 simultaneously, but the frequencies of the applied excitation are different from each other, and the data acquisition unit 30 acquires the voltage information of the second site and the third site (i.e., each electrode 104) with respect to the reference position and performs filtering processing to acquire the voltage information of the second site and the third site (i.e., each electrode 104) with respect to the reference position in all excitation states; or alternatively, the first and second heat exchangers may be,

the excitation control unit 20 applies excitation to the at least three electrode patches 102 simultaneously, but the frequencies of the applied excitation are different from each other, the data acquisition unit 30 acquires voltage information of the second site and the third site (i.e., each electrode 104) with respect to a reference position, and the processor unit 40 performs filtering processing on the voltage information of the second site and the third site (i.e., each electrode 104) acquired by the data acquisition unit 30 with respect to the reference position to acquire voltage information of the second site and the third site (i.e., each electrode 104) with respect to the reference position in all excitation states.

In the embodiment illustrated in fig. 1, there are 6 electrode patches 102 located at the back (H), groin (F), chest (C), back (B), left underarm (L), and right underarm (R), respectively, with the electrode patches 102 at the back (B) and the back (H) shown in dashed boxes. The excitation control unit 20 may select each of the two electrode patches 102 to apply excitation, so that there are 15 excitation states, which may be referred to as 15 excitation axes: RL, HF, CB, CR, CL, CH, CF, RB, LB, HB, FB, RF, LF, HR and HL. The excitation control unit 20 is not limited to apply excitation to each of the two electrode patches 102, and may apply excitation to N pairs of electrode patches 102 to generate an N (n.gtoreq.3) axis electric field in the target region of the target object. In practice, there may be a small current flowing through the target area by the excitation applied by one or more excitation axes, resulting in the electrode 104 in the target area not measuring the voltage value in the one or more excitation states or the measured voltage value in the one or more excitation states being significantly lower, for which case the excitation control unit 20 may discard the one or more excitation states (inactive excitation states) and not use them in subsequent calculations. The protection scope of the present invention should cover the case that when one or more excitation states are discarded, the different excitation states involved in the calculation include M effective excitation states selected from all the excitation states, M is not less than 3, N and M are natural numbers, and when all the excitation states are effective through the selection, N can be equal to M. Of course, if the stimulus states are not filtered, the different stimulus states involved in the calculation may also include all of the stimulus states.

In order to obtain the voltage information of each electrode 104 relative to the reference position in all the excitation states at the same sampling time, two implementation methods of time division or frequency division can be adopted. The time sharing method comprises the following steps: i.e. a continuously cycled high-speed switching between 15 excitations, the voltage values of each electrode 104 relative to the reference position are acquired in all excitation states. When the switching frequency is sufficiently high, it can be considered that the voltage value of each electrode 104 with respect to the reference position in all excitation states is acquired at the same time. The switching frequency may be 200kHz or 300kHz. The frequency dividing method comprises the following steps: i.e. signals of different frequencies are applied to the 15 excitation axes respectively, and the signals acquired at the electrodes 104 are then filtered by the data acquisition unit 30 or the processor unit 40 to obtain voltage values in all excitation states. For example, constant current/constant voltage signals of 10kHz, 10.5kHz, 11kHz, 11.5kHz …, etc. may be applied, respectively. Preferably, a frequency division method is used.

The data acquisition unit 30 is used for processing the transmitted data of all the sensors in the positioning device, including the magnetic field sensor 103 and the voltage sensor (i.e. the electrode 104). Amplification, filtering, analog-to-digital conversion, etc. are generally performed. Only with respect to the positioning function to be implemented in the present invention, the data to be collected by the data collecting unit 30 at each sampling instant is:

1. Magnetic field strength data at each magnetic field sensor 103;

2. voltage data for each electrode 104 (including electrode 104S and electrode 104T) relative to a reference position at all excitation states.

The positioning device further comprises a communication control unit 50 for connecting the processor unit 40 with the excitation control unit 20, the data acquisition unit 30 and the magnetic field generating unit 10, so as to control the communication and data transmission between the processor unit 40 and the magnetic field generating unit 10, the excitation control unit 20 and the data acquisition unit 30. The communication control unit 50 is a "transfer station" for signal and data transmission, and the transmission modes can be serial port, serial peripheral interface (Serial Peripheral Interface, SPI), I ² C. Wired modes such as a network port, USB and the like, and wireless methods such as wifi, bluetooth and the like can be selected. In addition, the communication control unit 50 may be used to connect other devices, such as a radio frequency ablation device, a CT device, an MRI device, an electrocardiograph, and the like, which are commonly used in operations, in addition to the unit modules in the positioning device.

Processor unit 40, typically a general purpose computer, is the core for system control and data processing. It can control the operation of the system by means of the communication control unit 40, in particular with the following functions:

1. The operation mode of the start-stop and excitation control unit of the magnetic field generation unit 10 is controlled;

2. the control data acquisition unit 30 synchronously acquires the magnetic field intensity of each magnetic field sensor 103 and the voltage value of each electrode 104 relative to the reference position;

3. based on the voltage value at the electrode 104S and the position coordinates of the electrode 104S, a neural network model (Neural network model of voltage-to-position, VP model, described above) describing the mapping relationship between the voltage value and the position coordinates is created;

4. the spatial position of the electrode 104T is calculated based on the VP model and the voltage value at the electrode 104T.

The processor unit 40 typically uses a software programming form to perform the functions described above. The software can be downloaded into a computer or used by a network. In addition, a memory may be provided in the processor unit 40 for storing the sensor data and the VP model related data.

In addition, the positioning device preferably further comprises a display unit 108, typically a display, which may be a CRT or LCD. A display unit 108 is communicatively connected to the processor unit 40 for displaying a position, a direction, a shape and/or a movement trajectory of the application tool of the interventional device 105 in the target object 100, wherein the position, the direction, the shape and/or the movement trajectory of the application tool of the interventional device 105 in the target object 100 is determined by the processor unit 40 based on the positioning result of the third site. It may be appreciated that a three-dimensional spatial model of the target object 100 may be pre-constructed during surgery, so that after the spatial position of the third site is located, the application catheter may be simulated according to the spatial position of the third site, so as to determine the position, direction and shape of the application catheter in the target object 100, further, since the position, direction and shape of the application catheter in the target object 100 are determined at each sampling time, the motion track of the application catheter in the target object 100 may be further determined. A three-dimensional body cavity model 109 is shown in the display unit 108 in fig. 1, as well as an end model 110 of the application catheter in the body cavity, which is drawn based on the positioning result of the processor unit 40. Of course, in other embodiments, the display unit 108 may also be configured to display a position, a direction, a shape and/or the motion trajectory of a learning tool in the target object 100, where the position, the direction, the shape and/or the motion trajectory of the learning tool in the target object 100 are determined by the processor unit 40 according to the positioning result of the first site and the second site.

Fig. 2 schematically shows a flow chart of the use of the positioning device according to the invention.

Step 200: at the beginning of the operation, a magnetic field and a multiaxial electric field are applied to the human body. The magnetic field is generally applied by the magnetic field generating unit 10, and a multi-axis electric field is generated by applying different electric current excitation to the plurality of electrode patches 102 and applying different electric current excitation to each of the patches on the surface of the target object 100, and either constant current excitation or constant voltage excitation can be selected.

The next process involves two phases, a training phase in which the learning catheter is operated and a positioning phase in which the application catheter is operated. The learning catheter and the application catheter are merely functionally differentiated catheters, in fact the same catheter may be used for both the training phase and the localization phase. The training phase starts earlier than the positioning phase, but allows both phases to operate at the same time for part of the time period.

Step 201: the learning catheter is placed in the heart chamber to walk around for a period of time.

During the training phase, the learning catheter may begin after being delivered into the target region 101 (e.g., heart chamber) through a vascular access. The learning catheter is integrated with a magnetic field sensor 103 and an electrode 104S, the distance between the two is fixed, and the distance information is known, so that the position information of the electrode 104S is obtained through calculation through the position and direction information of the magnetic field sensor 103 and the distance information of the two.

Step 202: during this time, the following data were measured continuously and synchronously:

1. the magnetic field strength information at the magnetic field sensor 103 on the catheter is learned. The magnetic field intensity information is used to calculate the spatial position and direction information of the magnetic field sensor 103 at each moment after being sent to the processor unit 40;

2. the voltage information at the catheter upper electrode 104S relative to some reference position is learned. The reference location will typically be selected to be a relatively fixed location, either on the body surface or at a location within the heart chamber (typically in the coronary sinus). The voltage is responsive to the aforementioned various excitation states.

Step 203: while collecting the data, the processor unit 40 continuously determines whether sufficient data has been accumulated based on the subsequent computational requirements. If the amount of data is insufficient, the roaming state of the catheter needs to be continued until the amount of data is sufficient, after which the operator is prompted by the display unit 108 to stop the training phase. The judging method can choose to set a fixed training time, and can also reduce the error to a certain degree by continuously training the neural network and then prompt the stopping, which will be described later.

Step 204: after sufficient data is accumulated, the positioning phase begins. The application catheter is placed into the heart chamber and the voltage at each electrode 104T on the application catheter is measured relative to a reference position.

Step 205: the processor unit 40 calculates the spatial position of each electrode 104T at each time according to the voltage value at each electrode 104T at each time, and the specific calculation method will be developed later.

It should be noted that the reference position and the respective electrode patch 102 positions for applying the excitation must remain relatively stable throughout the application, including the training phase and the positioning phase. If there is some manipulation in between (e.g., patient turning over) that results in displacement of electrode patch 102 beyond a certain threshold, the training phase must be restarted.

From the above description, the basic idea of the positioning method of the present invention is as follows:

1. applying a magnetic field generated by the magnetic field generating unit 10 and a multiaxial electric field to the body, the electric field being generated by applying constant current excitation/constant voltage excitation to the electrode patches 102 of the body surface;

2. training the voltage value measured at the electrode 104S and the position coordinate value of the electrode 104S (the position coordinate of the electrode can be indirectly calculated through the space position and direction information of the magnetic field sensor 103) by using a neural network model, and establishing a neural network model describing the mapping relation between the voltage value and the position coordinate value;

3. voltage data at the electrode 104T to be located is measured, and the position coordinates of the electrode 104T to be located are calculated using a neural network model.

Fig. 3 depicts one embodiment of the above-described positioning method:

step 300: a learning catheter equipped with a magnetic field sensor 103 and an electrode 104S is placed in the heart chamber and roams for a period of time; acquiring magnetic field strength data at the magnetic field sensor 103 and voltage data at the electrode 104S at each time; calculating a position and a direction at the magnetic field sensor 103 from the magnetic field intensity data at the magnetic field sensor 103; the position of the electrode 104S is calculated from the position and direction of the magnetic field sensor 103 and the distance of the magnetic field sensor 103 from the electrode 104S. Specifically, it can be according to formula P ₂ ＝P ₁ +D ₁ ·d ₁ Calculating the spatial position of the electrode 104S (i.e. the second site), wherein P ₂ Representing the spatial position of the electrode 104S (i.e., the second site), P ₁ And D ₁ Respectively the spatial position and direction of the magnetic field sensor 103 (i.e. the first site) (the direction being directed by the magnetic field sensor 103 towards the electrode 104S), d ₁ Representing the spatial distance between the magnetic field sensor 103 and the electrode 104S.

Step 301: assuming that at some point electrode 104S measures voltage V at a location P, a set of voltage-location combinations is referred to as a V-P data pair; the V-P data pair accumulated at the electrode 104S is used as a training sample to participate in neural network training, and a VP model describing the mapping relation between the voltage value measured at a certain position and the spatial coordinate value of the position in the target area to be positioned is constructed.

Step 302: an application catheter equipped with electrodes 104T is placed into the heart chamber, and the voltage value at each electrode 104T is measured.

Step 303: using the VP model obtained in step 301 and the voltage value measured in step 302, a positioning calculation of each electrode 104T is performed.

In a preferred embodiment, a fourth site and a fifth site may be included on the application catheter, such that during a localization phase, the data acquisition unit 30 may also acquire magnetic field strength information of the fourth site, voltage information of the fifth site relative to the reference position in all excitation states simultaneously. Correspondingly, the processor unit 40 calculates the spatial position information of the fifth locus according to the spatial position and direction information of the fourth locus and the spatial distance between the fourth locus and the fifth locus at each sampling time, and forms second class V-P data pairs by the voltage information of the fifth locus relative to the reference position and the spatial position information of the fifth locus under different excitation states, wherein the second class V-P data pairs can also participate in the training of the VP model.

For example, the learning catheter and the application catheter are the same catheter, that is, the fourth site and the first site are actually the same site, and the fifth site and the second site are actually the same site, so that the processor unit 40 may continuously acquire new V-P data pairs (i.e., the second V-P data pairs) during the positioning stage, and may also participate in the training of the VP model. Alternatively, the application catheter is not the same catheter as the learning catheter, but the magnetic field sensor 103 and the fifth site mount electrode 104S are mounted at the fourth site on the application catheter, so that the processor unit 40 will also continuously acquire new V-P data pairs (i.e., the second type V-P data pairs) during the positioning phase.

For the continuously acquired second class V-P data pairs, there may be the following three processing methods, preferably using the third:

a. maintaining the VP model trained in the previous training stage unchanged, namely maintaining the initial VP model unchanged, and not retraining the initial VP model according to the second V-P model so as to reduce calculation cost;

b. and updating the initial VP model according to the second V-P data pair to obtain a new VP model, namely, the first V-P data pair accumulated in the previous training stage and the second V-P data pair with the increasingly-increased positioning stage participate in training the VP model together. Thus, the data pairs participating in training are accumulated continuously, and the VP model is updated continuously, so that the latest VP model can be used for positioning calculation;

c. and training a neural network model according to the second V-P data to obtain a new VP model, namely, in the same way as the training stage, retraining by using the second V-P data obtained in the positioning stage to obtain the new VP model. More preferably, the training of the neural network model may be performed periodically according to the second type V-P data within a last preset time period to obtain a new VP model, where the preset time periods may be set according to actual situations, for example, set to 2 minutes. I.e. training is performed using only pairs of V-P data of the second type accumulated for a period of time before the current moment, i.e. the data used for training resembles a first-in first-out data queue. In this way, considering that the data acquisition environment is inevitably slowly changed along with the time, model training by using only the data in a period of time before the current moment can avoid that the accuracy of the VP model is influenced by the data in the period of time before the current moment. Thus, the VP model is updated continuously when the data pair participating in training is updated continuously, so that the latest VP model can be used for positioning calculation.

It will be appreciated that for the V-P data pair participating in the VP model training in the above embodiment, the voltage data V is directly measured by the electrode 104S, and the position data P is calculated based on the position of the magnetic field sensor 103 and the actual distance information between the magnetic field sensor and the electrode 104S, so that the VP model trained based on these two data can be considered to describe the real correspondence between these two data. This is emphasized because, in addition to the reliable position data provided at the electrode 104S with a known distance from the magnetic field sensor 103, a corresponding V-P data pair is generated after the positioning is completed at the electrode 104T to be positioned, but the position information of the electrode 104T at this time is calculated by the positioning method according to the present invention, and thus there may be some error.

In another embodiment, the processor unit 40 may further form a third V-P data pair from the voltage information of the third position relative to the reference position and the spatial position of the third position under different excitation states, screen out a valid data pair from the third V-P data pair, and update the initial VP model according to the valid data pair, so as to obtain a new VP model. For example, the positioning result based on the VP model at the electrode 104T may again form new V-P data pairs, and after a certain filtering condition is performed on the data pairs (for example, the V-P data pairs at the electrode 104T may also be added to training of the neural network, if the training error of the neural network can be reduced, the valid data pairs are valid, otherwise, the invalid data pairs) the valid data pairs may also participate in the training update of the initial VP model.

By the method, the spatial coordinates of any position (only by measuring the voltage data at the any position) can be calculated based on the information of the limited position (the position where the learning catheter roams) in the target area (such as the heart chamber) to be positioned, so that the positioning and tracking of the device (the catheter) to be positioned in the heart chamber can be completed.

The neural network type used for training the VP model can be an error back propagation neural network (Back Propagation Neural Network, hereinafter referred to as BP network), a radial basis function (Radial Basis Function, hereinafter referred to as RBF) neural network, or a support vector machine (Support Vector Machine, hereinafter referred to as SVM) neural network. Preferably, a BP network is used.

The above positioning method will be further explained using a BP network, taking a ring catheter as an example.

FIG. 4a shows an adjustable bend loop catheter having a shaft and a loop configuration; fig. 4b shows its sensor distribution structure, the catheter being used as both a learning catheter and a positioning catheter. Two electrodes 104S1 and 104S2 (i.e. the second site) and a magnetic field sensor 103 (i.e. the first site) are distributed on the rod, another 10 electrodes, denoted 104T1, 104T2 … … (i.e. the third site) are distributed on the ring, the spatial distance between the magnetic field sensor 103 and the electrodes 104S1 and 104S2 is fixed And are known, respectively denoted as d ₁ And d ₂ . With the positioning device of the invention, the positioning of the other electrodes (10 electrodes distributed on the ring) can be achieved by limited known information (the position of the magnetic field sensor 103 and the spatial distance of the magnetic field sensor 103 from the electrodes 104S1 and 104S 2).

The catheter was advanced into the right atrium via a vascular access, a ten-pole catheter was placed in the coronary sinus, and the upper electrode was used as the reference electrode, keeping the ten-pole catheter from moving throughout the procedure.

With the embodiment of fig. 1, electrode patches 102 are attached at 6 locations on the body surface, and excitation is applied using the combination of every two electrode patches 102 as one excitation axis, so that there are 15 excitation axes in total.

During the training phase, the catheter tip was maneuvered around the right atrium by operating the handle 106 for about 1min, and the following data were collected at each sampling instant:

1. magnetic field strength at the magnetic field sensor 103;

2. the two electrodes on the rod plus ten electrodes on the ring total 12 electrodes, voltage information at 15 excitation states relative to the reference position.

The processor unit 40 will calculate its position coordinates and direction vector based on the magnetic field strength at the magnetic field sensor 103. At a certain sampling time t, the spatial position and unit direction vector of the magnetic field sensor 103 in fig. 4b can be calculated from the magnetic field intensity information measured by the magnetic field sensor 103, assuming that the spatial position of the magnetic field sensor 103 is denoted as P ₁₀₃ The unit direction vector is denoted as D ₁₀₃ The direction is pointed 104S2 by electrode 104S 1.

At this time, the position P of the electrode 104S1 can be calculated _104S1 The method comprises the following steps: p (P) _104S1 ＝P ₁₀₃ -D ₁₀₃ ·d _104S1-103 ；

Position P of electrode 104S2 _104S2 The method comprises the following steps: p (P) _104S2 ＝P ₁₀₃ +D ₁₀₃ ·d _103-104S2 。

Meanwhile, the voltages measured at the electrodes 104S1 and 104S2 are U _104S1 And U _104S2 . Note that eachThe voltages are all 15 x 1 vectors, corresponding to the voltage values in 15 excitation states.

Thus, two sets of V-P data pairs are available at each sampling instant. Next, we describe how to train the V-P data pairs accumulated during the training phase using the BP network to construct the initial VP model.

Fig. 5 shows a structural diagram of a BP neural network used in the present embodiment, which is composed of the following three layers:

1. input layer: the voltage data measured at the lower electrode 104S at each sampling time is taken as an input layer of the BP network (the number of neurons of the input layer is determined by the number of effective excitation states, here 15, corresponding to the voltage values in 15 excitation states);

2. hidden layer: the number of hidden neurons is not too large or too small to describe the degree of non-linear distortion of the electric field, too large resulting in an excessive complexity of the calculation, typically 9-15 neurons, preferably 9 are chosen. The hidden layer transfer function is preferably a Sigmoid function;

3. Output layer: the spatial coordinate data at the lower electrode 104S at each sampling time (the time when the corresponding voltage data is acquired) is used as the output layer of the BP network, that is, the number of neurons of the output layer is fixed to 3 (corresponding to three spatial coordinate values of x, y and z). The output layer transfer function is preferably a Sigmoid function.

The following describes the training process of the network in more detail:

step one: and initializing a network. Determining the number of network neurons of each layer, wherein the number comprises 15 input layer neurons, 9 hidden layer neurons and 3 output layer neurons; initializing weights among layers of the network, including a weight w between an input layer and a hidden layer _ij And weight w between hidden layer and output layer _jk The method comprises the steps of carrying out a first treatment on the surface of the Initializing threshold b of hidden layer _j And output layer threshold b _k The method comprises the steps of carrying out a first treatment on the surface of the Where i represents the input layer i-th neuron (i=1, 2 …), j represents the hidden layer j-th neuron (j=1, 2 …), and k represents the output layer k-th neuron (k=1, 2, 3). Each weight and threshold is initialized to a random number between-1 and 1. In addition, the hidden layer transfer function and the output layer transfer function are determined to be Sigmoid functions.

Step two: and calculating the output value of each layer of network in turn. Let the input voltage vector be v ₁ ,v ₂ …v ₁₅ Then the output h of the j-th neuron of the hidden layer _j The method comprises the following steps:

output p of the kth neuron of the output layer _k The method comprises the following steps:

where f () represents a transfer function between layers, i.e., sigmoid function.

Step three: calculating the output mean square error E of the current training sample as an objective function of the BP network:

wherein p is _k For the calculated coordinate values of the electrode 104S that are the current output values of the neural network,

is the target output value of the neural network, i.e., the actual coordinate value of the electrode 104S.

Step four: based on Gradient Descent (Gradient) strategy, each parameter in the network, namely the weight w of each network layer, is subjected to negative Gradient direction of objective function _ij 、w _jk And threshold b _j 、b _k Is adjusted. The calculation of the weight and threshold values for adjusting the BP network using the gradient descent method is a general method and will not be described in detail here.

Step five: and (3) continuously iterating the training process from the second step to the fourth step until the preset training times or the expected error size are reached, and stopping the iterative process to finish the training of the BP network.

Thus, the construction of the initial VP model, that is, the mapping relationship between the voltage value measured at the electrode 104S and the position of the electrode 104S, is completed.

Then, in the positioning stage, the voltage values of each electrode 104T in all excitation states at each moment are continuously collected, and the spatial coordinate values of each electrode 104T can be obtained by using the initial VP model (i.e., performing the calculation in the step two above), so as to complete the positioning of the annular catheter in the heart chamber. Fig. 4c shows the configuration of the annular duct on the display unit 108 at a certain point in time calculated using this positioning method.

It can be appreciated that in the current three-dimensional electrocardiographic mapping field, high-density mapping analysis is an increasingly important development trend. While the implementation of high density mapping is not separated from high precision catheter positioning. In the current catheter positioning technology, the positioning technology based on the magnetic field has higher precision but higher cost. The common electrode used for electric field positioning can acquire data required by a positioning function while finishing mapping, and can effectively reduce cost. However, the nonlinear distortion of the electric field in the body is strong, so that the accuracy is difficult to ensure. Therefore, most of the magnetoelectric double positioning technology is applied in the market to control the cost and improve the positioning precision. However, the magnetoelectric dual-localization technology on the market is mostly based on local linear assumption of electric field, and a linear model is still used for modeling in-vivo electric field. This can lead to either insufficient positioning accuracy or to extreme complexity of the linear model used. The invention uses the neural network technology to generate the nonlinear transformation relation model of the electric field and the space position, thereby ensuring enough positioning precision and effectively reducing the calculated amount.

It should be noted that the catheter shown in fig. 4a is only used for clarity of description, and in fact, the present invention is not limited to the shape shown in the drawings, but can be applied to catheters of other forms such as a balloon or basket. In the example, a BP network is used to train to obtain a VP model, but any method is used to establish a mapping relationship model between voltage data (measured by a certain electrode and corresponding to a certain excitation state) and position data (space where the electrode is located) by using a neural network method to realize electrode positioning, which is within the scope of protection of the patent.

In addition, although the embodiments described herein are mostly applied to electrode positioning in the heart chamber, they may be applied in other situations, such as neurologic surgery, tumor ablation, pulmonary artery angioplasty, etc., where invasive diagnosis or treatment with interventional or implantable devices is required.

From the above description, in the practical application process of the invention, the positioning of the common catheter electrode can be realized after the catheter with the magnetic field sensor enters the heart chamber to walk for a period of time. Considering nonlinear distortion effects of in-vivo electric fields, most of positioning methods in the prior art are "learning" based on a certain local linearity assumption, however, not only the positioning accuracy is insufficient, but also more compensation or calibration calculation is needed to meet the requirements of overall nonlinearity and local linearity, and further the calculation amount is increased. The invention is different from the prior art in that the local linear assumption of the electric field is abandoned, and the nonlinear conversion model between the electric field and the spatial position is constructed by using the nonlinear fitting characteristic of the neural network technology so as to solve the problems. Therefore, the positioning accuracy can be ensured, and the calculated amount can be effectively reduced.

Based on the same inventive concept, the present invention also provides an interventional procedure system comprising a positioning device of an interventional device as described above and the interventional device. Wherein the interventional device may be a multi-limb catheter or a ring catheter.

the method comprises the following steps:

Preferably, the neural network model comprises an input layer, a plurality of hidden layers and an output layer which are sequentially connected, the number of neurons of the input layer is equal to the number of different excitation states, and the number of neurons of the output layer is equal to 3.

Preferably, the initial VP model is trained as follows:

Preferably, the application device further comprises a fourth site and a fifth site;

the method further comprises the steps of:

Preferably, the application device and the learning device are the same device, the fourth site and the first site are the same site, and the fifth site and the second site are the same site.

Preferably, the new VP model is trained as follows:

Preferably, the training a neural network model according to the second class V-P data to obtain a new VP model includes:

Preferably, the method further comprises:

Preferably, the neural network model is selected from one of an error back propagation neural network, a radial basis function neural network and a support vector machine neural network.

Preferably, the method further comprises:

Preferably, the first site is provided with a magnetic field sensor, and the data acquisition unit acquires magnetic field intensity information of the first site through the magnetic field sensor;

Preferably, the different excitation states comprise M effective excitation states selected from all the excitation states, wherein M is more than or equal to 3 and less than or equal to N; or (b)

The different excitation states include all of the excitation states.

Based on the same inventive concept, the present invention also provides a computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, implements a method of positioning an interventional device as described above. The storage medium may include: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

Based on the same inventive concept, the invention also provides a computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the method of positioning an interventional device as described above when executing the computer program.

Specifically, in the embodiment of the present invention, the processor may be a central processing unit (centralprocessing unit, abbreviated as CPU), and the processor may also be other general purpose processors, digital signal processors (digital signalprocessor, abbreviated as DSP), application specific integrated circuits (application specificintegrated circuit, abbreviated as ASIC), off-the-shelf programmable gate arrays (field programmable gate array, abbreviated as FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and so on. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

It should also be appreciated that the memory in embodiments of the present invention may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The nonvolatile memory may be a read-only memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an electrically Erasable EPROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (random accessmemory, RAM for short) which acts as an external cache. By way of example but not limitation, many forms of random access memory (random access memory, abbreviated as RAM) are available, such as static random access memory (static RAM), dynamic Random Access Memory (DRAM), synchronous Dynamic Random Access Memory (SDRAM), double data rate synchronous dynamic random access memory (double datarate SDRAM, abbreviated as DDR SDRAM), enhanced Synchronous Dynamic Random Access Memory (ESDRAM), synchronous Link DRAM (SLDRAM), and direct memory bus random access memory (direct rambus RAM, abbreviated as DR RAM).

In this document, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The above description is only illustrative of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention, and any alterations and modifications made by those skilled in the art based on the above disclosure shall fall within the scope of the appended claims.

Claims

1. A positioning apparatus of an interventional device, characterized in that the interventional device comprises a learning means and an application means, the learning means and the application means being adapted to be placed in a target area of a target object;

the processor unit is configured to calculate, at each sampling time, spatial position information of the second site according to spatial position and direction information of the first site and spatial distance between the first site and the second site, and form a first type V-P data pair from voltage information of the second site relative to the reference position and spatial position information of the second site in different excitation states, where the first type V-P data pair is used to train a neural network model to obtain an initial VP model for describing a mapping relationship between voltage and position, where the spatial position and direction information of the first site are calculated according to magnetic field strength information of the first site; and in a positioning stage, calculating the spatial position of the third site by using the initial VP model according to the voltage information of the third site relative to the reference position under different excitation states at each sampling time;

The neural network model comprises an input layer, a plurality of hidden layers and an output layer which are sequentially connected, wherein the number of neurons of the input layer is equal to the number of different excitation states, and the number of neurons of the output layer is equal to 3;

taking the voltage information in the first V-P data pair as the input of the neural network model, and calculating the spatial position information predicted by the output layer; calculating an objective function value of the neural network model according to the spatial position information in the first V-P data pair and the spatial position information predicted by the output layer; and optimizing model parameters of the neural network model according to the objective function value until a preset training ending condition is met, so as to obtain the initial VP model.

2. The positioning apparatus of claim 1 wherein said applicator further comprises fourth and fifth sites thereon;

3. The positioning apparatus of claim 2, wherein the application means is the same means as the learning means, the fourth site is the same site as the first site, and the fifth site is the same site as the second site.

4. The positioning device of claim 2, wherein the processor unit is configured to update the initial VP model based on the second V-P data pair to obtain a new VP model.

5. The positioning device of claim 2, wherein the processor unit is configured to train a neural network model based on the second type V-P data to obtain a new VP model.

6. The positioning device of claim 5, wherein the processor unit is configured to periodically train a neural network model based on the second type V-P data within a last predetermined time period to obtain a new VP model.

7. A positioning device as claimed in any one of claims 1-3, characterized in that the processor unit is further configured to combine the voltage information of the third position relative to the reference position and the spatial position information of the third position in different excitation states into a third class V-P data pair, to screen out valid data pairs from the third class V-P data pair, and to update the initial VP model based on the valid data pairs, to obtain a new VP model.

8. The positioning device of any of claims 1-6, wherein the neural network model is selected from one of an error back propagation neural network, a radial basis function neural network, a support vector machine neural network.

9. The positioning device of any of claims 1-6, wherein the excitation control unit applies excitation to the at least three electrode patches and switches at a high speed that is continuously cycled between all excitation states, the data acquisition unit acquiring voltage information of the second and third sites relative to the reference position in all excitation states; or alternatively, the first and second heat exchangers may be,

10. The positioning device of any of claims 1-6, wherein the excitation is constant current excitation or constant voltage excitation.

11. The positioning device of any of claims 1-6, further comprising a communication control unit for connecting the processor unit and the magnetic field generating unit, the excitation control unit, the data acquisition unit to control communication and data transmission between the processor unit and the magnetic field generating unit, the excitation control unit, the data acquisition unit.

12. The positioning device of any of claims 1-6, further comprising a display unit communicatively coupled to the processor unit for displaying a position, a direction, a shape, and/or a motion profile of the application implement in the target object, wherein the position, the shape, the direction, and/or the motion profile of the application implement in the target object are determined by the processor unit based on a result of positioning the third site.

13. The positioning device of any of claims 1-6, wherein the first site is provided with a magnetic field sensor, and the data acquisition unit acquires magnetic field strength information of the first site through the magnetic field sensor;

14. Positioning device according to any of the claims 1-6, wherein,

the different excitation states comprise M effective excitation states which are screened from all the excitation states, wherein M is more than or equal to 3 and less than or equal to N; or (b)

The different excitation states include all of the excitation states.

15. An interventional procedure system, characterized by comprising a positioning device of an interventional device according to any of claims 1-14 and the interventional device.

16. A method of positioning an interventional device, characterized in that the interventional device comprises a learning means and an application means, the learning means and the application means being adapted to be placed in a target area of a target object;

the method comprises the following steps:

In a positioning stage, receiving voltage information of a third site on the application tool relative to the reference position in all excitation states, which is synchronously acquired at each sampling time, and calculating the spatial position of the third site by using the initial VP model according to the voltage information of the third site relative to the reference position in different excitation states;

17. The positioning method of claim 16, wherein the application tool further comprises a fourth site and a fifth site thereon;

The method further comprises the steps of:

18. The positioning method of claim 17 wherein the application means is the same means as the learning means, the fourth site is the same site as the first site, and the fifth site is the same site as the second site.

19. The positioning method according to claim 17, characterized in that a new VP model is trained in the following way:

20. The positioning method according to claim 17, characterized in that a new VP model is trained in the following way:

21. The positioning method according to claim 20, wherein training a neural network model according to the second class V-P data to obtain a new VP model comprises:

22. The positioning method according to any one of claims 16-18, wherein the method further comprises:

23. The positioning method according to any of claims 16-21, wherein the neural network model is selected from one of an error back propagation neural network, a radial basis function neural network, a support vector machine neural network.

24. The positioning method according to any one of claims 16-21, further comprising:

25. The positioning method according to any one of claims 16-21, wherein a magnetic field sensor is provided at the first site, and a data acquisition unit acquires magnetic field strength information of the first site through the magnetic field sensor;

26. The positioning method according to any one of claims 16-21,

The different excitation states include all of the excitation states.

27. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any of claims 16-26.

28. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 16-26 when the computer program is executed.